Will It Run AI
black forest labs

Flux.2 Klein 4B

Frontier

by Black Forest Labs

Lightweight 4B variant of FLUX.2 for efficient generation. Distilled from FLUX.2-dev for faster inference on consumer GPUs. Apache 2.0 licensed — the most accessible Flux model for commercial use.

VRAM requirements, GPU fit, and setup notes for Flux.2 Klein 4B, including 8GB/12GB fit guidance where relevant. Recommended runtimes: ComfyUI and Diffusers support. Best download size: ~4.0 GB at FP8.

  • Only 4B params — runs on 8GB+ GPUs
  • Apache 2.0 — fully open for commercial use
  • Distilled from FLUX.2-dev
  • Great quality-to-speed ratio
HuggingFaceDocumentation
271K downloads574 likes
ComfyUI, DiffusersFP8 safetensors

Your hardware

Detecting...

Parameters4B
Max Resolution1024×1024
Default Steps20
ArchitectureDIT
Licenseapache-2.0

Image Quality Benchmarks

Measured quality metrics for Flux.2 Klein 4B outputs.

Human Preference Score82%

How often humans prefer this model's output (0-100%)

Aesthetic Score7.5

Visual quality and composition rating (5-9 scale)

VRAM Requirements by Resolution and Precision

Compare which GPUs can run Flux.2 Klein 4B at different precisions. FP8 uses less memory than FP16 when available, and the grade shows how comfortably each GPU handles the workload.

FP16 (full precision)

ResolutionVRAM RequiredRTX 4090 24GBRTX 3060 12GBRTX 4060 8GBMacBook Pro M4 Pro 24GB
512×5129.8 GBSADS
768×7689.9 GBSADS
1024×102410.0 GBSADS

FP8 (~40% less VRAM)

ResolutionVRAM RequiredRTX 4090 24GBRTX 3060 12GBRTX 4060 8GBMacBook Pro M4 Pro 24GB
512×51210.1 GBSADS
768×76810.3 GBSADS
1024×102410.6 GBSADS

Optimization Tips

Turbo / LCM distillation

Use distilled scheduler at 4-8 steps for faster iteration

Run with Python

Run with Python (diffusers)
from diffusers import FluxPipeline
import torch

pipe = FluxPipeline.from_pretrained(
    "black-forest-labs/FLUX.2-klein-4B",
    torch_dtype=torch.float16
)
pipe.to("cuda")

image = pipe(
    prompt="your prompt here",
    num_inference_steps=20,
    guidance_scale=3.5,
    height=1024,
    width=1024,
).images[0]
image.save("output.png")

Get started

Setup instructions for running Flux.2 Klein 4B locally

1. Download the model

Get the checkpoint from HuggingFace

2. Place in:

ComfyUI/models/checkpoints/

3. Launch ComfyUI

python main.py

Memory Breakdown

VRAM allocation at 1024×1024 on RTX 4090 24GB (24 GB)

Required: 10.0 GBAvailable: 24.0 GB
Weights8.0 GB
VAE0.2 GB
Text Encoder9.6 GB
Activations0.6 GB
Overhead0.5 GB

Estimated Generation Time

Time per image at 1024×1024, 28 steps, FP16.

RTX 4090 24GB~1.2s
RTX 3060 12GB~10.2s
RTX 4060 8GB~6.8s
MacBook Pro M4 Pro 24GB~21.9s

Sample Outputs

Available Formats, Downloads & Setup

Download Flux.2 Klein 4B in the precision that matches your GPU. Lower precision usually means less VRAM pressure, while higher precision keeps more quality.

格式精度大小提供商
safetensors推荐FP168.0 GBofficial下载
safetensorsFP84.0 GBofficial下载

LoRA Ecosystem

Limited

Limited LoRA availability. Some FLUX.2 LoRAs may be compatible.

Related Workflows

You might also like

Frequently asked questions

FAQ — Flux.2 Klein 4B VRAM, Runtimes & Fit

How much VRAM does Flux.2 Klein 4B need?

Flux.2 Klein 4B (4B parameters) requires approximately 10.0 GB of VRAM at FP16 precision for standard 1024×1024 image generation. If you want a lighter setup, lower precisions like FP8 can reduce memory use when available.

Can I run Flux.2 Klein 4B on an 8GB GPU?

Flux.2 Klein 4B usually needs more than 8GB for comfortable local use. Check the VRAM table above for the exact resolution and precision trade-off.

Does Flux.2 Klein 4B work in ComfyUI and Diffusers?

Flux.2 Klein 4B is marked for ComfyUI and Diffusers support in our catalog, so those are the runtimes we recommend first for local setup. If your workflow uses another front end, check the model's available formats and workflow notes above before downloading.

Can I run Flux.2 Klein 4B on RTX 4090?

Yes, the RTX 4090 (24 GB VRAM) can run Flux.2 Klein 4B comfortably at FP16. Expected generation time is around ~1.2s per image at 1024×1024.

Does Flux.2 Klein 4B support ControlNet?

There are currently no known ControlNet adapters for Flux.2 Klein 4B. Check Hugging Face and Civitai for community-contributed adapters.

Does Flux.2 Klein 4B have LoRA support?

Limited LoRA availability. Some FLUX.2 LoRAs may be compatible. The LoRA ecosystem for Flux.2 Klein 4B is rated as "minimal". Each LoRA adds roughly 0.2 GB of extra VRAM.

How fast is Flux.2 Klein 4B?

On a reference GPU (RTX 4090 24GB), Flux.2 Klein 4B generates a 1024×1024 image in approximately ~1.2s at FP16 with 28 inference steps. Faster GPUs with higher memory bandwidth will produce images more quickly.

About Flux.2 Klein 4B

Use cases
photorealisticfast-generationlightweight
Recommended runtimes
comfyuidiffusers

See also